README.md

---
license: mit
license_name: non-commercial-license-dyna1
license_link: https://github.com/WaymentSteeleLab/Dyna-1/blob/main/LICENSE.txt
tags:
- proteins
- nucleic acids
pretty_name: NAKB
size_categories:
- n<1K
viewer: true
configs:
- config_name: main
  data_files:
  - split: DNA
    path: DNA/DNA.csv
  - split: Hybrid
    path: Hybrid/Hybrid.csv
  - split: PNA
    path: PNA/PNA.csv
  - split: Protein
    path: Protein/Protein.csv
  - split: RNA
    path: RNA/RNA.csv
  - split: other
    path: other/other.csv
---
<h1>Original Paper:</h1>

<p>Lawson CL, Berman HM, Vallat B, Chen L, Zirbel C (2024) The Nucleic Acid Knowledgebase: a new portal for 3D structural information about nucleic acids. Nucleic Acids Research 52, D245-D254. </p>
<p>https://doi.org/10.1093/nar/gkad957</p>

<h1>Nucleic Acid Knowledgebase (NAKB)</h1>
<p>NAKB data set contains 21166 structures including Nucleic Acids, Protein, and Ligand Annotations, and determined 3D structures found in the Nucleic Acid Database (NDB) and the Protein Data Bank (PDB), including structures determined by X-ray, Electron Microscopy, and Nuclear Magnetic Resonance. Information contained in this data set is PDB/NDB ID, Release Date, Sequence, Polymer Composition, and Molecular Weight. Assemblies, split by chain with sequences added to the curated data of Lawson et al. </p>

<h1>Quickstart Usage</h1>
<strong>Install HuggingFace Datasets package</strong>
<p>Each subset can be loaded into python using the Huggingface datasets library. First, from the command line install the datasets library:</p>

```bash
$ pip install datasets
```

<p>then, from within python load the datasets library</p>

```python
import datasets
from datasets import load_dataset
```

<strong>Load model datasets</strong>
<p>To load one of the NAKB_sub model datasets:</p>

```python
from datasets import load_dataset

dataset_protein = load_dataset(
    "RosettaCommons/Rosetta_NAKB",
    data_files="Protein/*.csv"
)

dataset_protein["train"].head()
```

<h2>Data Explaination</h2>
<p>Each row in the dataset represents a single polymer chain within a larger macromolecular assembly. The dataset includes the following key fields:</p>

* `polyclass`: The overall assembly classification. This describes the type of macromolecular complex in which the chain is found (e.g., Protein/RNA). It reflects the biological composition of the full structure rather than the individual molecule.

* `released`: Structure release date.

* `resolution`: Experimental resolution in Ångströms.

* `method`: Structure determination method (e.g., EM for electron microscopy).

* `chains`: Chain identifier within the structure.

* `polytype`: The polymer type of the individual chain (e.g., Protein, RNA). Unlike polyclass, which describes the entire assembly, polytype refers specifically to the molecular identity of the single chain represented in that row.

* `description`: Functional or canonical name of the molecule (e.g., ribosomal proteins such as 40S or 60S subunits).

* `entityMW`: Molecular weight of the entity (kDa).

* `seq`: Amino acid or nucleotide sequence of the polymer chain.

<h2><code>polyclass</code> vs <code>polytype</code> </h2>

<p>It is important to distinguish between the <code>polyclass</code> and <code>polytype</code>attributes.</p>

<p><code>polyclass</code> defines the biological assembly category (e.g., a ribonucleoprotein complex composed of proteins and RNA).</p>

<p><code>polytype</code> defines the chemical nature of the specific chain listed in that row.</p>
<p>For example, in a Protein/RNA polyclass assembly (such as a ribosome), individual rows may have a polytype of Protein (e.g., ribosomal proteins) or RNA (e.g., rRNA chains). Thus, multiple polytypes can exist within a single polyclass. This distinction allows the dataset to represent both the system-level composition of a structure and the molecular identity of its individual components.</p>